Figure 4 - uploaded by Karl Fritscher
Content may be subject to copyright.
Comparison of the DXA images (top) and the projection of the BMD distribution from the 3D reconstruction (DRRs, mid- dle). Subtraction between the pairs of images (bottom). Represent- ative examples of the mean BMD distribution accuracy (Table 1).
Source publication
The diagnosis of osteoporosis and the prevention of femur fractures is a major challenge for our society. However, the diagnosis performed in clinical routine from Dual Energy X-ray Absorptiometry (DXA) images is limited. This paper proposes a 3D reconstruction method of both the shape and the Bone Mineral Density (BMD) distribution of the proximal...
Contexts in source publication
Context 1
... Radiographs (DRRs) generated from the model. Similar to [10], the deformation of the shape model needs to be applied to the BMD distribution model inside the shape. This was achieved by applying a Thin Plate Spline transformation defined by the deformation of the shape. The similarity criterion to be optimized is the average of the mean absolute error obtained between each pair of DXA image and DRR (Figure 2). A database of 64 specimens of human proximal femurs (all female, with a mean age of 80 ± 10 years) was collected for a previous study [15] from the institute of Anatomy at the Ludwig Maximi- lians University Munich (Germany). All these bones were scanned using a 16-row MSCT scanner (Sensation 16; Siemens Medical Solutions, Erlangen, Germany). All the CT-scans were resampled from a spatial resolution of 0.195*0.195*0.5mm 3 to a spatial resolution of 0.5*0.5*0.5mm 3 and calibrated using a phantom to obtain a QCT analysis. This database was divided into a first database of 44 samples (mean age: 80 ± 10 years) for the construction of the statistical atlas (see section 2.1) and a second database of 20 samples (mean age: 81 ± 10 years) for the evaluation of the 3D reconstruction method (see section 2.2). Among these 20 samples, 10 were defined as osteoporotic and 10 as non-osteoporotic based on the WHO criteria [1] (aBMD measurements had been performed from frontal DXA images of the specimens, however, these DXA images have not been stored). Since the DXA image acquisitions of these specimens were not available, simulated DXA images were generated from the QCT volumes using a ray-casting technique (resolution of 0.3*0.3mm 2 ). This technique allowed the generation of realistic DXA images corresponding to true DXA images [16]. This allowed us to easily investigate five different configurations, in terms of the number of simulated DXA images and view angles (enumerated in Table 1). For each configuration, the 3D reconstructions obtained from a single or multi-view simulated DXA images were compared with the QCT volumes, which were regarded as the "ground truth". To evaluate the shape accuracy, each of the 20 QCT volumes was semi-automatically segmented using ITK-SNAP [11] (as done for the reference shape built in section 2.1). The shapes obtained from the 3D reconstruction method were subsequently superimposed (Iterative Closest Point method [17]) onto their corresponding segmentation and the point-to-surface distances were computed. In order to estimate the accuracy of the BMD distribution, the recon- structed volume was aligned to the ground truth CT volume using the transformation resulting from the previous Iterative Closest Point registration and the BMD differences were estimated at each voxel. The mean shape accuracy, in comparison with the "ground truth" QCT segmentations, was 1.3mm from one view and between 0.8mm and 0.9mm from 2, 3 or 4 views (Table 1). Error maps showing the distribution of the mean shape differences are pro- vided in Figure 3. The average BMD distribution accuracy, resulting from the voxel by voxel comparison between the reconstructions and the "ground truth" QCT volumes was 81mg/cm 3 from one view and between 53 and 60mg/cm 3 from 2, 3 or 4 views (Table 1). In comparison to the mean range of values in terms of density observed in the QCT volumes (1856mg/cm 3 ), this mean difference represents errors of 4.4% (one view) and between 2.9 and 3.2% (2, 3 or 4 views). This study aimed at proposing a 3D reconstruction method of both femur shape and BMD distribution from DXA images. Several studies were performed to recover the femoral shape from DXA images [6, 8, 9]. To our knowledge, the method presented in this paper is the first one that allows a 3D reconstruction of the BMD distribution, and evaluates the accuracy of the method. In this context, several configurations (single and multi-view DXA images) were investigated. From only one frontal view, the 3D reconstructions were found quite accurate for both the shape (mean error: 1.3mm) and the BMD distribution (mean error: 4.4%) (Table 1). The comparison between the frontal DXA image and the DRR (Figure 4, "1 view") highlighted that the projection of the BMD distribution in the 3D reconstruction is consistent in comparison to the DXA image. Regarding the multi-view configurations, the addition of the second view (sagittal) resulted in a gain of accuracy for both the shape (mean error: 0.9mm) and the BMD distribution (mean error: 3.2%) (Table 1 and Figure 3). This shape accuracy is similar to the one evaluated in [8] (mean shape accuracy of 0.8mm using a 3D reconstruction method from two DXA images; frontal and sagittal). However, this method, limited to the shape reconstruction, required an operator time of 10 minutes, dedicated to manual adjustments of the model [8]. In comparison, our method is fully automated and non-supervised. The addition of other views (3 or 4 view configurations) brings a slight gain in terms of accuracy. Figure 4 high- lights the consistency between the projections of the BMD distribution and the DXA images (configuration "3 views.2"). Note that the two configuration "3 views.1" and "3 views.2" were equivalent in terms of accuracy. By relying on the interesting results that we obtained in this in vitro context, this approach is currently evaluated for in vivo clinical applications. Although the evaluation presented in this paper was performed from simulated DXA images, the simulation technique used has been shown to produce realistic simulated DXA images from QCT [16]. Consequently, to apply this method to "true" DXA images is not a major issue. However, with clinical DXA images of patients, we need to deal with the superimposition of other bony structures such as the pelvis. A preliminary evaluation performed from DXA images of patients have recently con- firmed that this method provides accurate 3D reconstructions, even in this in vivo clinical context. The method proposed in this paper allows the 3D reconstruction of the proximal femur from DXA images with a satisfactory accuracy in terms of shape and BMD distribution. From one DXA image, this method is compatible with the current clinical practice, since most of the clinical sites are equipped with single-view DXA imaging devices. For the multi-view DXA medical systems equipped with a C-arm, that are appearing in clinics, the configuration "2 views" (frontal and sagittal) yield the best compromise. This better characterization of the femoral bone, from clinical routine imaging devices, is expected to provide a better diagnosis of osteoporosis and, consequently, a better prevention of femur fractures. We are currently investigating the potential of this method for such in vivo clinical applications. This research has been financially supported by the 3D-FemOs project (3-Dimensional Femur Reconstruction for Osteoporotic Fracture Risk Assessment) with a grant awarded by ACC1Ó Valtec (Tipus 1). In addition this work has been supported by a grant from the Instituto de Salud Carlos III (FI09/00757) and by a grant from the "Deutsche Forschungsgemeinschaft" (LO 730 / 3-1). The au- thors acknowledge Benedikt Schuler (Institute for Biomedical Image Analysis, UMIT, Hall in Tirol, Austria) and Dr. E. M. Lochmüller (Universitäts-Frauenklinik der LMU, München, Germany) for the data ...
Context 2
... are used to deform each volume to the same mean reference shape using Thin Plate Spline interpolation [14]. PCA is then applied to the bone density volumes resulting in a BMD distribution model of femur. To sum up, the statistical model is thus described by a set of parameters defining the shape and a set of parameters characteriz- ing the BMD distribution (Figure 1). Our recent work [7] allows us to acquire a 3D reconstruction (shape and BMD distribution), using the atlas described above, from a single-view DXA image. In this paper, we extend this method to the context of multi-view DXA images. The 3D reconstruction is achieved by searching the parameter space of the statistical models (together with a translation, rotation and uniform scal- ing) that maximizes the similarity between the DXA images and the Digitally Reconstructed Radiographs (DRRs) generated from the model. Similar to [10], the deformation of the shape model needs to be applied to the BMD distribution model inside the shape. This was achieved by applying a Thin Plate Spline transformation defined by the deformation of the shape. The similarity criterion to be optimized is the average of the mean absolute error obtained between each pair of DXA image and DRR (Figure 2). A database of 64 specimens of human proximal femurs (all female, with a mean age of 80 ± 10 years) was collected for a previous study [15] from the institute of Anatomy at the Ludwig Maximi- lians University Munich (Germany). All these bones were scanned using a 16-row MSCT scanner (Sensation 16; Siemens Medical Solutions, Erlangen, Germany). All the CT-scans were resampled from a spatial resolution of 0.195*0.195*0.5mm 3 to a spatial resolution of 0.5*0.5*0.5mm 3 and calibrated using a phantom to obtain a QCT analysis. This database was divided into a first database of 44 samples (mean age: 80 ± 10 years) for the construction of the statistical atlas (see section 2.1) and a second database of 20 samples (mean age: 81 ± 10 years) for the evaluation of the 3D reconstruction method (see section 2.2). Among these 20 samples, 10 were defined as osteoporotic and 10 as non-osteoporotic based on the WHO criteria [1] (aBMD measurements had been performed from frontal DXA images of the specimens, however, these DXA images have not been stored). Since the DXA image acquisitions of these specimens were not available, simulated DXA images were generated from the QCT volumes using a ray-casting technique (resolution of 0.3*0.3mm 2 ). This technique allowed the generation of realistic DXA images corresponding to true DXA images [16]. This allowed us to easily investigate five different configurations, in terms of the number of simulated DXA images and view angles (enumerated in Table 1). For each configuration, the 3D reconstructions obtained from a single or multi-view simulated DXA images were compared with the QCT volumes, which were regarded as the "ground truth". To evaluate the shape accuracy, each of the 20 QCT volumes was semi-automatically segmented using ITK-SNAP [11] (as done for the reference shape built in section 2.1). The shapes obtained from the 3D reconstruction method were subsequently superimposed (Iterative Closest Point method [17]) onto their corresponding segmentation and the point-to-surface distances were computed. In order to estimate the accuracy of the BMD distribution, the recon- structed volume was aligned to the ground truth CT volume using the transformation resulting from the previous Iterative Closest Point registration and the BMD differences were estimated at each voxel. The mean shape accuracy, in comparison with the "ground truth" QCT segmentations, was 1.3mm from one view and between 0.8mm and 0.9mm from 2, 3 or 4 views (Table 1). Error maps showing the distribution of the mean shape differences are pro- vided in Figure 3. The average BMD distribution accuracy, resulting from the voxel by voxel comparison between the reconstructions and the "ground truth" QCT volumes was 81mg/cm 3 from one view and between 53 and 60mg/cm 3 from 2, 3 or 4 views (Table 1). In comparison to the mean range of values in terms of density observed in the QCT volumes (1856mg/cm 3 ), this mean difference represents errors of 4.4% (one view) and between 2.9 and 3.2% (2, 3 or 4 views). This study aimed at proposing a 3D reconstruction method of both femur shape and BMD distribution from DXA images. Several studies were performed to recover the femoral shape from DXA images [6, 8, 9]. To our knowledge, the method presented in this paper is the first one that allows a 3D reconstruction of the BMD distribution, and evaluates the accuracy of the method. In this context, several configurations (single and multi-view DXA images) were investigated. From only one frontal view, the 3D reconstructions were found quite accurate for both the shape (mean error: 1.3mm) and the BMD distribution (mean error: 4.4%) (Table 1). The comparison between the frontal DXA image and the DRR (Figure 4, "1 view") highlighted that the projection of the BMD distribution in the 3D reconstruction is consistent in comparison to the DXA image. Regarding the multi-view configurations, the addition of the second view (sagittal) resulted in a gain of accuracy for both the shape (mean error: 0.9mm) and the BMD distribution (mean error: 3.2%) (Table 1 and Figure 3). This shape accuracy is similar to the one evaluated in [8] (mean shape accuracy of 0.8mm using a 3D reconstruction method from two DXA images; frontal and sagittal). However, this method, limited to the shape reconstruction, required an operator time of 10 minutes, dedicated to manual adjustments of the model [8]. In comparison, our method is fully automated and non-supervised. The addition of other views (3 or 4 view configurations) brings a slight gain in terms of accuracy. Figure 4 high- lights the consistency between the projections of the BMD distribution and the DXA images (configuration "3 views.2"). Note that the two configuration "3 views.1" and "3 views.2" were equivalent in terms of accuracy. By relying on the interesting results that we obtained in this in vitro context, this approach is currently evaluated for in vivo clinical applications. Although the evaluation presented in this paper was performed from simulated DXA images, the simulation technique used has been shown to produce realistic simulated DXA images from QCT [16]. Consequently, to apply this method to "true" DXA images is not a major issue. However, with clinical DXA images of patients, we need to deal with the superimposition of other bony structures such as the pelvis. A preliminary evaluation performed from DXA images of patients have recently con- firmed that this method provides accurate 3D reconstructions, even in this in vivo clinical context. The method proposed in this paper allows the 3D reconstruction of the proximal femur from DXA images with a satisfactory accuracy in terms of shape and BMD distribution. From one DXA image, this method is compatible with the current clinical practice, since most of the clinical sites are equipped with single-view DXA imaging devices. For the multi-view DXA medical systems equipped with a C-arm, that are appearing in clinics, the configuration "2 views" (frontal and sagittal) yield the best compromise. This better characterization of the femoral bone, from clinical routine imaging devices, is expected to provide a better diagnosis of osteoporosis and, consequently, a better prevention of femur fractures. We are currently investigating the potential of this method for such in vivo clinical applications. This research has been financially supported by the 3D-FemOs project (3-Dimensional Femur Reconstruction for Osteoporotic Fracture Risk Assessment) with a grant awarded by ACC1Ó Valtec (Tipus 1). In addition this work has been supported by a grant from the Instituto de Salud Carlos III (FI09/00757) and by a grant from the "Deutsche Forschungsgemeinschaft" (LO 730 / 3-1). The au- thors acknowledge Benedikt Schuler (Institute for Biomedical Image Analysis, UMIT, Hall in Tirol, Austria) and Dr. E. M. Lochmüller (Universitäts-Frauenklinik der LMU, München, Germany) for the data ...
Similar publications
This study aims to investigate patterns of femoral cortical bone fragility with age (at death) and to evaluate its associations with sex and bone mineral density. Radiogrammetric parameters of the femur and bone mineral density at the proximal femur were assessed in an adult sample (N=98) from the Coimbra Identified Skeletal Collection (Portugal)....
Differences in skeletal geometry may generate different patterns of mechanical loading to bone. Impact and muscle loading during physical activity have been shown to influence skeletal geometry. The purpose of this study was to compare geometric measures of the pelvis and proximal femur (PF) of young children and to analyze the contribution and pot...
Osteoporosis is a multi factorial disease with dimension of genetic and nutritional considerations. The aim of this study was to present data from the association of plasma zinc, copper and toxic elements of lead and cadmium levels with bone mineral density in Iranian women. 135 women gave their information and enrolled. Fasting plasma was used for...
Background
Rheumatoid arthritis (RA) is associated with poor bone mineral density (BMD). We designed the current study owing to the lack of long-term prospective studies regarding whether a high disease activity leads to increased bone loss.
Methods
We have continually enrolled patients with RA. According to the average disease activity score in 2...
In a multi-center patient study, using different CT scanners, CT-based finite element (FE) models are utilized to calculate failure loads of femora with metastases. Previous studies showed that using different CT scanners can result in different outcomes. This study aims to quantify the effects of 1) different CT scanners, 2) different CT protocols...
Citations
... Additionally, no prior works were found to have developed DXA-based statistical appearance/intensity models for the proximal tibia. Instead, studies have typically focused on the proximal femur to assist with the diagnosis of osteoporosis [33]. Depending on the level of accuracy needed to achieve successful patient outcomes, another alternative could be to use a statistical model based on demographic information (such as age, sex, and ethnicity) to estimate required implant stiffness. ...
In knee arthroplasty, implants are used to replace the articulating surfaces of the tibia and femur bones, with most constituting of solid metallic components. Consequentially, biomechanical stresses and strains are no longer adequately distributed at the joint post-surgery, preventing beneficial bone remodeling. To mitigate this studies have explored additively manufacturing implants with porous lattice structures to match the mechanical properties of bone. Authors have also outlined how such structures can be designed using computed tomography data to simulate the stiffness of individuals’ bones. Such methods however currently require substantial manual work by trained professionals to process the image files, extract the density information, and design lattice structures. This study proposes what is believed to be the first fully automatic pipeline capable of producing tibial trays with compliant structures customized specifically for individuals’ bones, achieved using machine learning methods. The novel process, combining classification, object detection, and segmentation machine learning models, used to facilitate the automated workflow, is outlined. The efficaciousness of the pipeline is then demonstrated by testing it using clinical computed tomography data and comparing the results with those obtained manually. As a proof of concept, prototype designs generated by the pipeline with differing degrees of complexity, up to and including mapping stiffness variation in 3D through the shaft of the tibia, were also fabricated.
... DEXA represents a prognosis of the mineral content. Some works are done to construct 3D distribution of BMD from DEXA images for better assessment of osteoporosis [56][57][58]. ...
... Humbert et al. [57], proposed a 3D reconstruction method of both the femoral shape and the 3D BMD distribution from DXA images. The reconstruction accuracy acquired from single-view and multi-view DEXA devices was measured. ...
... Ahmad et al. [56] "deformable" 2D/3D registration of statistical atlases Femur 48 subjects The correlation between QCT (quantitative computed tomography) derived BMD and the vBMD from VXA (Volumetric DXA) is 0.98. Ludovic Humbert et al. [57] Based on statistical atlas Femur 20 bone specimens BMD accuracy of 4.4% from a single-view DEXA image and 3.2% from multi-view DEXA image. Tristan Whitmarsh et al. [58] intensity based 3D-2D registration process ...
Osteoporosis is one of the common bone diseases that reduces bone strength and affects the structure of the bone and thereby increases the chances of fracture risk, more likely in the spine, hip, and wrist. The diagnosis of osteoporosis is done by measuring the bone quality and bone mass, mainly bone mineral density (BMD). There are various methods available to measure the BMD. Of which, Dual Energy X-ray Absorptiometry (DEXA) is considered as the benchmark method. BMD is a parameter to determine the important score known as T-score that determines the osteoporosis condition. BMD measurement is visualized in X-ray images and DEXA images. This survey article focuses on the measurement of BMD using various benchmark image processing algorithms such as image enhancement, segmentation, and texture analysis on X-ray and DEXA images for osteoporosis detection. Superior properties of DEXA uncover the prospective for new medical applications and researches. The article reviews the early methods of BMD measurements in a nutshell. Also, the article mainly explains the features that are on par with the X-ray and DEXA images. The article explains the image processing algorithms used for osteoporosis detection. The methods such as pre-processing, feature extraction, and segmentation methods are explained for both X-ray and DEXA imaging modalities.
... The first strategy seeks to establish correspondence, and minimize the geometric distance, between features detected on both the deformable model and the patient's X-ray im- age(s). This is accomplished using a Kriging optimization (also known as a Gaussian process regression) [4], [9], [16], [17], [19], [20], [22], [23], [25], [26], [28], [30]- [32], [35]- [37], [40]- [44], [46], [48]- [51], [54], [62]. It is however first necessary to identify these corresponding features. ...
... These, however, only evaluate shape information. In order to gauge the accuracy of the reconstructed intensity information, voxel- wise comparisons have been proposed [47], [51], [62]. In fact any intensity-based similarity metrics, such as those discussed in Section II-C, can be extended to 3D and used. ...
... They exploited two parameters, namely the image scale (in mm/pixel) and the distance from the COP to the detector (the imaging plane), which can be retrieved from a DICOM image (provided that the anatomical structure was imaged using a standard clinical procedure). The researchers in [47], [62], [140] and [55] have reported success when using a single image of actual patients, and also reported on the accuracy of the reconstructed shape and intensity information, but only when using DXA. Dual-energy X- ray images provides two advantages over conventional X-ray images. ...
Patient-specific three-dimensional (3D) bone models are useful for a number of clinical applications such as surgery planning, postoperative evaluation as well as implant and prosthesis design. Two--dimensional--to--three--dimensional (2D/3D) reconstruction, also known as model-to-modality or atlas--based 2D/3D registration, provides a means of obtaining a 3D model of a patient's bone(s) from their 2D radiographs, when 3D imaging modalities are not available. The preferred approach to estimating both shape and density information (that would be present in a patient's CT data) for 2D/3D reconstruction makes use of digitally reconstructed radiographs and deformable models in an iterative, non-rigid, intensity-based approach. Based on a large number of state-of-the-art 2D/3D bone reconstruction methods, a unified mathematical formulation of the problem is proposed in a common conceptual framework, using unambiguous terminology. In addition, shortcomings, recent adaptations and persisting challenges are discussed along with insights for future research.
... In addition, CT scanning is not a routine imaging modality for assessing fracture risk and it also exposes the patient to additional radiation dose compared to DXA (Damilakis et al., 2010). Researchers have looked at generating 3D FE models from 2D DXA images with some success (Humbert et al., 2010;Thevenot et al., 2014;Väänänen et al., 2015), but there are still significant errors in the prediction of geometry and the assigned material properties from a single image. ...
Osteoporosis and related bone fractures are an increasing global burden in our ageing society. Areal bone mineral density assessed through dual energy X-ray absorptiometry (DEXA), the clinically accepted and most used method, is not sufficient to assess fracture risk individually. Finite element (FE) modelling has shown improvements in prediction of fracture risk, better than aBMD from DEXA, but is not practical for widespread clinical use. The aim of this study was to develop an adaptive neural network (ANN)-based surrogate model to predict femoral neck strains and fracture loads obtained from a previously developed population-based FE model. The surrogate model performance was assessed in simulating two loading conditions: the stance phase of gait and a fall.
The surrogate model successfully predicted strains estimated by FE (r2= 0.90-0.98 for level gait load case, r2= 0.92-0.95 for the fall load case). Moreover, an ANN model based on three measurements obtainable in clinics (femoral neck length (level gait) or maximum femoral neck diameter (fall), femoral neck bone mass, body weight) was able to give reasonable predictions (r2= 0.84-0.94) for all of the strain metrics and the predicted femoral neck fracture load. Overall, the surrogate model has potential for clinical applications as they are based on simple measures of geometry and bone mass which can be derived from DEXA images, accurately predicting FE model outcomes, with advantages over FE models as they are quicker and easier to perform.
... For example, rotations and translations of training images with respect to each other are eliminated during the alignment procedure. There are certain algorithms for automated alignment that are described in the supplementary material (Section S2.1.2) among which Generalized Procrustes Analysis (GPA) [8][9][10][11][12] is the most widely used algorithm. ...
... These techniques include DXA, quantitative ultrasound (QUS), quantitative computed tomography (QCT), and digital x-ray radiogrammetry [25]. The most widely used way of assessing bone mass in clinical practice is based on aBMD measures derived from DXA scans [11,26,27]. However, low BMD measures alone are not sufficient to explain all osteoporotic fractures [28]. ...
... There are certain algorithms for automated alignment that are described in the supplementary material (Section S2.1.2) among which Generalized Procrustes Analysis (GPA) [8][9][10][11][12] is the most widely used algorithm. This method is efficient and simple to implement and is therefore expected to remain popular for the alignment of the training shapes. ...
... Currently, most medical images are organized, and the images are created using software. It has been mentioned that 2-D measurements of a DXA image were not accurate [18]. As a result, some medical images were manually traced by performing segmentation in order to improve their accuracy. ...
... Software was used to join the data during the segmentation process, which is a very time-consuming process. The ITK-SNAP software provides a semiautomated segmentation algorithm [18]. The result from this software can be refined by the operator, but the user is still required to control the medical field reconstruction methods when using this software. ...
... Refer to [19] for a survey related to 3-D reconstructions of medical images. The authors of [18] proposed a 3-D reconstruction method for femur shapes to determine bone mineral density from DXA images. This method was performed by searching the parameter space of the statistical model that contained the highest similary of DXA images and digitally reconstructed radiographs (DDRs) generated by the model. ...
Surface reconstruction by using 3-D data is used to represent the surface of an object and perform important tasks. The type of data used is important and can be described as either structured or unstructured. For unstructured data, there is no connectivity information between data points. As a result, incorrect shapes will be obtained during the imaging process. Therefore, the data should be reorganized by finding the correct topology so that the correct shape can be obtained. Previous studies have shown that the Kohonen self-organizing map (KSOM) could be used to solve data organizing problems. However, 2-D Kohonen maps are limited because they are unable to cover the whole surface of closed 3-D surface data. Furthermore, the neurons inside the 3-D KSOM structure should be removed in order to create a correct wireframe model. This is because only the outside neurons are used to represent the surface of an object. The aim of this paper is to use KSOM to organize unstructured data for closed surfaces. KSOM isused in this paper by testing its ability to organize medical image data because KSOM is mostly used in constructing engineering field data. Enhancements are added to the model by introducing class number and the index vector, and new equations are created. Various grid sizes and maximum iterations are tested in the experiments. Based on the results, the number of redundancies is found to be directly proportional to the grid size. When we increase the maximum iterations, the surface of the image becomes smoother. An area formula is used and manual calculations are performed to validate the results. This model is implemented and images are created using Dev C++ and GNUPlot.
... Several methods that are able to estimate the 3D femoral shape based on two or several radiographs have been developed ( Laporte et al. 2003;Kolta et al. 2005;Filippi et al. 2007;Kurazume et al. 2009). Use of only one radiograph and a shape template has also been introduced to estimate the 3D shape of a femur ( Langton et al. 2009b;Väänänen et al. 2011;Humbert et al. 2010;Galibarov et al. 2010). Only two of these studies were able to estimate both shape and the internal volumetric BMD (vBMD) inside the femur ( Humbert et al. 2010;Galibarov et al. 2010). ...
... Use of only one radiograph and a shape template has also been introduced to estimate the 3D shape of a femur ( Langton et al. 2009b;Väänänen et al. 2011;Humbert et al. 2010;Galibarov et al. 2010). Only two of these studies were able to estimate both shape and the internal volumetric BMD (vBMD) inside the femur ( Humbert et al. 2010;Galibarov et al. 2010). These methods are based on deforming a template to the shape of a 2D X-ray, by fitting the surface of the template to the contours of the X-ray image directly or by applying statistical appearance models. ...
... Earlier, others have reported larger errors (Galibarov et al.: mean distance differences 2.57-3.74 mm ( Galibarov et al. 2010); Humbert et al.: mean distance difference 1.3 mm ( Humbert et al. 2010)). Previous studies have suggested that the voxel size in the CT images may affect the estimation accuracy ( Langton et al. 2009b;Väänänen et al. 2011). ...
Measurement of bone mineral density (BMD) by dual-energy X-ray absorptiometry (DXA) alone is only a moderate predictor of fracture risk. Finite element analysis (FEA) of bone mechanics, based on DXA images, may improve the prediction of fracture risk. We developed a method to estimate the 3D shape and density distribution of the proximal femur, using a 2D BMD image and a femur shape template. Proximal femurs of eighteen human cadavers were imaged using computed tomography and divided into two sets (N = 9 + 9). The template was created from the samples in first set by using 3D generalized Procrustes analysis and thin-plate splines. Subsequently, the template and 2D BMD image were utilized to estimate the shape and internal density distribution of the femurs in the second set. Finally, FEA was conducted based on the original and the estimated bone models to evaluate the effect of geometrical and density distributional errors on the mechanical strength. The volumetric errors induced by the estimation itself were low (<1.4%). In the estimation of bones in the second set, the mean distance difference between the estimated and the original bone surfaces was 0.80 ± 0.19 mm, suggesting feasible estimation of the femoral shape. The mean absolute error in voxel-by-voxel BMD was 120±8 mg cm⁻³. In FEA, the stiffness of the proximal femur differed by -7±16% between the original and estimated bones. The present method, in comparison with methods used in previous studies, improved the prediction of the geometry, the BMD distribution and the mechanical characteristics of the proximal femur. Potentially, the proposed method could ultimately improve the determination of bone fracture risk.
... Methods to assess the 3D shape of the femur based on one radiograph and a shape template have been presented (Langton et al. 2009, Väänänen et al. 2010, Galibarov et al. 2010, Humbert et al. 2010). Only two of these studies were also able to estimate the internal architecture of the femur (Galibarov et al. 2010, Humbert et al. 2010). ...
... Methods to assess the 3D shape of the femur based on one radiograph and a shape template have been presented (Langton et al. 2009, Väänänen et al. 2010, Galibarov et al. 2010, Humbert et al. 2010). Only two of these studies were also able to estimate the internal architecture of the femur (Galibarov et al. 2010, Humbert et al. 2010). However, the accuracy of the internal architecture estimation was not evaluated in 3D. ...
Measurement of bone mineral density (BMD) by DXA (dual-energy X-ray absorptiometry) is generally considered to be the clinical golden standard technique to diagnose osteoporosis. However, BMD alone is only a moderate predictor of fracture risk. Finite element analyses of bone mechanics can contribute to a more accurate prediction of fracture risk. In this study, we applied a method to estimate the 3D geometrical shape of bone based on a 2D BMD image and a femur shape template. Proximal femurs of eighteen human cadavers were imaged with computed tomography (CT) and divided into two groups. Image data from the first group (N = 9) were applied to create a shape template by using the general Procrustes analysis and thin plate splines. This template was then applied to estimate the shape of the femurs in the second group (N = 9), using the 2D BMD image projected from the CT image, and the geometrical errors of the shape estimation method were evaluated. Finally, finite element analysis with stance loading condition was conducted based on the original CT and the estimated geometrical shape to evaluate the effect of the geometrical errors on the outcome of the simulations. The volumetric errors induced by the shape estimation method itself were low (<0.6%). Increasing the number of bone specimens used for the template decreased the geometrical errors. When nine bones were used for the template, the mean distance difference (±SD) between the estimated and the CT shape surfaces was 1.2 ± 0.3 mm, indicating that the method was feasible for estimating the shape of the proximal femur. Small errors in geometry led systematically to larger errors in the mechanical simulations. The method could provide more information of the mechanical characteristics of bone based on 2D BMD radiography and could ultimately lead to more sensitive diagnosis of osteoporosis.
The primary purpose of this paper is to outline a methodology for evaluating the likelihood of cortical bone fracture in the proximal femur in the event of a sideways fall. The approach includes conducting finite element (FE) analysis in which the cortical bone is treated as an anisotropic material, and the admissibility of the stress field is validated both in tension and compression regime. In assessing the onset of fracture, two methodologies are used, namely the Critical Plane approach and the Microstructure Tensor approach. The former is employed in the tension regime, while the latter governs the conditions at failure in compression. The propagation of localized damage is modeled using a constitutive law with embedded discontinuity (CLED). In this approach, the localized deformation is described by a homogenization procedure in which the average properties of cortical tissue intercepted by a macrocrack are established. The key material properties governing the conditions at failure are specified from a series of independent material tests conducted on cortical bone samples tested at different orientations relative to the loading direction. The numerical analysis deals with simulations of experiments involving the sideways fall, and the results are compared with the experimental data. This includes both the evolution of fracture and the local load-displacement characteristics. The proposed approach is numerically efficient, and the results do not display a pathological mesh-dependency. Also, in contrast to the XFEM approach, the analysis does not require any extra degrees of freedom.
SummaryA new technique to enhance hip fracture risk prediction in older adults was presented and assessed. The new method dramatically improved prediction at high specificity levels using only a standard clinical diagnostic scan. This has the potential to be implemented in clinical practice to enhance patient fragility diagnosis.IntroductionDiagnosis of osteoporosis is based on the measurement of bone mineral density (BMD) using dual-energy X-ray absorptiometry (DXA) scans. However, studies have shown this to be insufficient to accurately predict hip fractures. Therefore, complementary methods are needed to enhance hip fracture risk prediction to identify vulnerable patients.Methods
Hip DXA scans were obtained for 192 subjects from the Canadian Multicenter Osteoporosis Study (CaMos), 50 of whom had experienced a hip fracture within 5 years of the scan. 2D statistical shape and appearance modeling was performed to account for the effect of the femur’s geometry and BMD distribution on hip fracture risk. Statistical shape modeling (SSM), and statistical appearance modeling (SAM) were also used separately to predict the fracture risk based solely on the femur’s geometry and BMD distribution, respectively. Combined with BMD, age, and body mass index (BMI), logistic regression was performed to estimate the fracture risk over the 5-year period.ResultsUsing the new technique, hip fractures were correctly predicted in 78% of cases compared with 36% when using the T-score. The accuracy of the prediction was not greatly reduced when using SSM and SAM (78% and 74% correct, respectively). Various geometric and BMD distribution traits were identified in the fractured and non-fractured groups.Conclusion2D SSAM can dramatically improve hip fracture prediction at high specificity levels and estimate the year of the impending fracture using standard clinical images. This has the potential to be implemented in clinical practice to estimate hip fracture risk.
